Simulation-Based Model Verification and Performance Validation

kyuhan Kim

2025-08-09


This document outlines the simulation-based verification and performance testing of a newly developed length-based age-structured model applied to pollock in South Korean waters. This model serves as an operating model that generates simulated population dynamics rather than fitting to observed data. The analysis employs forward simulation experiments to serve as both a sanity check for model correctness and comprehensive validation across parameter configurations.

  • Equilibrium Benchmark Testing: Deterministic simulations verify that the model converges to analytically expected equilibrium states across different parameter combinations, ensuring correct implementation of the underlying population dynamics.

  • Relative Difference Analysis: Relative differences in abundance time series are computed for pairwise parameter changes (\(\gamma\), \(\eta\), \(\zeta\)) to assess model sensitivity and stability across fixed parameter combinations. This analysis identifies how individual parameter changes affect population dynamics.

  • Stochastic Recruitment Analysis: Recruitment dynamics are tested under different parameter configurations and autocorrelation coefficients of \(\rho_R = 0.0\) and \(\rho_R = 0.4\) to confirm marginal distribution consistency across models.


Equilibrium benchmark test

Relative difference time series for fixed pairs

Recruitment deviations and relative recruitment size